Key Considerations for Robotic Process Automation

Author: Spiros Alexiou, Ph.D., CISA, CSX-F, CIA
Date Published: 20 October 2020

Robotic Process Automation (RPA) is a modern trend in which a typically simple (i.e., requiring no real intelligence) sequence, that up until now involved a human operator, is replaced with a virtual robot. This decision is driven by cost-cutting considerations and perceived inherent computer advantages over humans with respect to precision and speed. However, there is no magical one-size-fits-all solution. RPA is not necessarily the best solution in all cases, and it comes with challenges and risk of its own.

The first question to ask is if there is a repetitive set of tasks, such as registering a sale, computing value-added tax (VAT) or sending an invoice, then why is it not already automated? The technology and tools to automate tasks have been available for more than 50 years and include daemons and interfaces. It is important to note that the tasks we are referring to are typically very simple tasks. You do not need a subject matter expert human operator or someone with very high intelligence to read a sale, send an invoice or compute VAT.

In these types of cases, the correct solution is to redesign the process and automate all the steps. For example, you can automate so that the output from the sale registration is fed to the VAT calculator and its output to a proper invoice form. If RPA is used in this situation, it instead repeats the steps the old way (e.g., the sale registration is shown on a computer screen; it is read from the screen to compute VAT, which is shown in another screen; an email is prepared to inform the invoice operator what to record, etc.). This solution does not make the most sense. If we have the data in the computer’s native (binary) form, why show them on the screen (which is a preferred form for a human operator, not a computer), re-read them and then digitize them?

If this does not make sense, why do it this way? One obvious answer is legacy applications. We repeat the same steps as in the legacy way, except that we often have the computer read the screen output, convert to digital form and proceed. Again, this is not the most efficient solution. Time saved may be very modest, specifically compared with a redesigned solution with communication. From an auditor point of view, RPA with reading from the screen may also be risky as misreading a character (such as a decimal point) may have grave financial consequences. When auditing such solutions, practitioners should keep these considerations in mind and assess the scope of each particular case.

Editor’s note: For further insights on this topic, read Spiros Alexiou’s recent Journal article, “The Dark Side of Robotic Process Automation,” ISACA Journal, volume 5, 2020.